Analysing Efficiency of New Load Distribution Algorithm with Horizontal Scaling

Abstract

As the distributed software and microservice architecture is spreading, the importance of load distribution is increasing. There are multiple indicators to measure the quality of the distribution, in the past the main interest was load balancing, i.e. evenly disrtibuting the load among available units. In this paper another aspect is investigated, how well can the total load be distributed among the least number of units, keeping the others in a gracefully shut down state. For this purpose an unevenly distributing algorithm is presented and compared to the model of traditional load distribution. The algorithm is designed to support horizontal scaling of processing units via attempting to operate most of the active units at a design load level, allowing only one at the time to take less traffic. This paper focuses on the specific case of load distribution inside telecommunication core networks.